Seongsik Park

Jongwan Kim, DongJin Lee, Byunggook Na et al.

Event cameras respond to brightness changes in the scene asynchronously and independently for every pixel. Due to the properties, these cameras have distinct features: high dynamic range (HDR), high temporal resolution, and low power consumption. However, the results of event cameras should be processed into an alternative representation for computer vision tasks. Also, they are usually noisy and cause poor performance in areas with few events. In recent years, numerous researchers have attempted to reconstruct videos from events. However, they do not provide good quality videos due to a lack of temporal information from irregular and discontinuous data. To overcome these difficulties, we introduce an E2V-SDE whose dynamics are governed in a latent space by Stochastic differential equations (SDE). Therefore, E2V-SDE can rapidly reconstruct images at arbitrary time steps and make realistic predictions on unseen data. In addition, we successfully adopted a variety of image composition techniques for improving image clarity and temporal consistency. By conducting extensive experiments on simulated and real-scene datasets, we verify that our model outperforms state-of-the-art approaches under various video reconstruction settings. In terms of image quality, the LPIPS score improves by up to 12% and the reconstruction speed is 87% higher than that of ET-Net.

Jiyong Moon, Seongsik Park

One of the key issues in facial expression recognition in the wild (FER-W) is that curating large-scale labeled facial images is challenging due to the inherent complexity and ambiguity of facial images. Therefore, in this paper, we propose a self-supervised simple facial landmark encoding (SimFLE) method that can learn effective encoding of facial landmarks, which are important features for improving the performance of FER-W, without expensive labels. Specifically, we introduce novel FaceMAE module for this purpose. FaceMAE reconstructs masked facial images with elaborately designed semantic masking. Unlike previous random masking, semantic masking is conducted based on channel information processed in the backbone, so rich semantics of channels can be explored. Additionally, the semantic masking process is fully trainable, enabling FaceMAE to guide the backbone to learn spatial details and contextual properties of fine-grained facial landmarks. Experimental results on several FER-W benchmarks prove that the proposed SimFLE is superior in facial landmark localization and noticeably improved performance compared to the supervised baseline and other self-supervised methods.

Jiyong Moon, Junseok Lee, Yunju Lee et al.

Recently, vision Transformers (ViTs) have been actively applied to fine-grained visual recognition (FGVR). ViT can effectively model the interdependencies between patch-divided object regions through an inherent self-attention mechanism. In addition, patch selection is used with ViT to remove redundant patch information and highlight the most discriminative object patches. However, existing ViT-based FGVR models are limited to single-scale processing, and their fixed receptive fields hinder representational richness and exacerbate vulnerability to scale variability. Therefore, we propose multi-scale patch selection (MSPS) to improve the multi-scale capabilities of existing ViT-based models. Specifically, MSPS selects salient patches of different scales at different stages of a multi-scale vision Transformer (MS-ViT). In addition, we introduce class token transfer (CTT) and multi-scale cross-attention (MSCA) to model cross-scale interactions between selected multi-scale patches and fully reflect them in model decisions. Compared to previous single-scale patch selection (SSPS), our proposed MSPS encourages richer object representations based on feature hierarchy and consistently improves performance from small-sized to large-sized objects. As a result, we propose M2Former, which outperforms CNN-/ViT-based models on several widely used FGVR benchmarks.

Dayena Jeong, Jaewoo Park, Jeonghee Jo et al.

Recent deep neural networks (DNNs), such as diffusion models [1], have faced high computational demands. Thus, spiking neural networks (SNNs) have attracted lots of attention as energy-efficient neural networks. However, conventional spiking neurons, such as leaky integrate-and-fire neurons, cannot accurately represent complex non-linear activation functions, such as Swish [2]. To approximate activation functions with spiking neurons, few spikes (FS) neurons were proposed [3], but the approximation performance was limited due to the lack of training methods considering the neurons. Thus, we propose tendency-based parameter initialization (TBPI) to enhance the approximation of activation function with FS neurons, exploiting temporal dependencies initializing the training parameters.

Byunggook Na, Jisoo Mok, Seongsik Park et al.

Spiking neural networks (SNNs) that mimic information transmission in the brain can energy-efficiently process spatio-temporal information through discrete and sparse spikes, thereby receiving considerable attention. To improve accuracy and energy efficiency of SNNs, most previous studies have focused solely on training methods, and the effect of architecture has rarely been studied. We investigate the design choices used in the previous studies in terms of the accuracy and number of spikes and figure out that they are not best-suited for SNNs. To further improve the accuracy and reduce the spikes generated by SNNs, we propose a spike-aware neural architecture search framework called AutoSNN. We define a search space consisting of architectures without undesirable design choices. To enable the spike-aware architecture search, we introduce a fitness that considers both the accuracy and number of spikes. AutoSNN successfully searches for SNN architectures that outperform hand-crafted SNNs in accuracy and energy efficiency. We thoroughly demonstrate the effectiveness of AutoSNN on various datasets including neuromorphic datasets.

Byunggook Na, Jaehee Jang, Seongsik Park et al.

Various deep learning applications on smartphones have been rapidly rising, but training deep neural networks (DNNs) has too large computational burden to be executed on a single smartphone. A portable cluster, which connects smartphones with a wireless network and supports parallel computation using them, can be a potential approach to resolve the issue. However, by our findings, the limitations of wireless communication restrict the cluster size to up to 30 smartphones. Such small-scale clusters have insufficient computational power to train DNNs from scratch. In this paper, we propose a scalable smartphone cluster enabling deep learning training by removing the portability to increase its computational efficiency. The cluster connects 138 Galaxy S10+ devices with a wired network using Ethernet. We implemented large-batch synchronous training of DNNs based on Caffe, a deep learning library. The smartphone cluster yielded 90% of the speed of a P100 when training ResNet-50, and approximately 43x speed-up of a V100 when training MobileNet-v1.

Seongsik Park, Harksoo Kim

Sentence-level relation extraction mainly aims to classify the relation between two entities in a sentence. The sentence-level relation extraction corpus often contains data that are difficult for the model to infer or noise data. In this paper, we propose a curriculum learning-based relation extraction model that splits data by difficulty and utilizes them for learning. In the experiments with the representative sentence-level relation extraction datasets, TACRED and Re-TACRED, the proposed method obtained an F1-score of 75.0% and 91.4% respectively, which are the state-of-the-art performance.

Dongjin Lee, Seongsik Park, Jongwan Kim et al.

Spiking neural networks (SNNs) have been gaining interest as energy-efficient alternatives of conventional artificial neural networks (ANNs) due to their event-driven computation. Considering the future deployment of SNN models to constrained neuromorphic devices, many studies have applied techniques originally used for ANN model compression, such as network quantization, pruning, and knowledge distillation, to SNNs. Among them, existing works on knowledge distillation reported accuracy improvements of student SNN model. However, analysis on energy efficiency, which is also an important feature of SNN, was absent. In this paper, we thoroughly analyze the performance of the distilled SNN model in terms of accuracy and energy efficiency. In the process, we observe a substantial increase in the number of spikes, leading to energy inefficiency, when using the conventional knowledge distillation methods. Based on this analysis, to achieve energy efficiency, we propose a novel knowledge distillation method with heterogeneous temperature parameters. We evaluate our method on two different datasets and show that the resulting SNN student satisfies both accuracy improvement and reduction of the number of spikes. On MNIST dataset, our proposed student SNN achieves up to 0.09% higher accuracy and produces 65% less spikes compared to the student SNN trained with conventional knowledge distillation method. We also compare the results with other SNN compression techniques and training methods.

Seongsik Park, Sungroh Yoon

The tremendous energy consumption of deep neural networks (DNNs) has become a serious problem in deep learning. Spiking neural networks (SNNs), which mimic the operations in the human brain, have been studied as prominent energy-efficient neural networks. Due to their event-driven and spatiotemporally sparse operations, SNNs show possibilities for energy-efficient processing. To unlock their potential, deep SNNs have adopted temporal coding such as time-to-first-spike (TTFS)coding, which represents the information between neurons by the first spike time. With TTFS coding, each neuron generates one spike at most, which leads to a significant improvement in energy efficiency. Several studies have successfully introduced TTFS coding in deep SNNs, but they showed restricted efficiency improvement owing to the lack of consideration for efficiency during training. To address the aforementioned issue, this paper presents training methods for energy-efficient deep SNNs with TTFS coding. We introduce a surrogate DNN model to train the deep SNN in a feasible time and analyze the effect of the temporal kernel on training performance and efficiency. Based on the investigation, we propose stochastically relaxed activation and initial value-based regularization for the temporal kernel parameters. In addition, to reduce the number of spikes even further, we present temporal kernel-aware batch normalization. With the proposed methods, we could achieve comparable training results with significantly reduced spikes, which could lead to energy-efficient deep SNNs.

Seongsik Park, Dongjin Lee, Sungroh Yoon

Spiking neural networks (SNNs) have emerged as energy-efficient neural networks with temporal information. SNNs have shown a superior efficiency on neuromorphic devices, but the devices are susceptible to noise, which hinders them from being applied in real-world applications. Several studies have increased noise robustness, but most of them considered neither deep SNNs nor temporal information. In this paper, we investigate the effect of noise on deep SNNs with various neural coding methods and present a noise-robust deep SNN with temporal information. With the proposed methods, we have achieved a deep SNN that is efficient and robust to spike deletion and jitter.

Seongsik Park, Harksoo Kim

Relation extraction is a type of information extraction task that recognizes semantic relationships between entities in a sentence. Many previous studies have focused on extracting only one semantic relation between two entities in a single sentence. However, multiple entities in a sentence are associated through various relations. To address this issue, we propose a relation extraction model based on a dual pointer network with a multi-head attention mechanism. The proposed model finds n-to-1 subject-object relations using a forward object decoder. Then, it finds 1-to-n subject-object relations using a backward subject decoder. Our experiments confirmed that the proposed model outperformed previous models, with an F1-score of 80.8% for the ACE-2005 corpus and an F1-score of 78.3% for the NYT corpus.

Seongsik Park, Seijoon Kim, Byunggook Na et al.

Spiking neural networks (SNNs) have gained considerable interest due to their energy-efficient characteristics, yet lack of a scalable training algorithm has restricted their applicability in practical machine learning problems. The deep neural network-to-SNN conversion approach has been widely studied to broaden the applicability of SNNs. Most previous studies, however, have not fully utilized spatio-temporal aspects of SNNs, which has led to inefficiency in terms of number of spikes and inference latency. In this paper, we present T2FSNN, which introduces the concept of time-to-first-spike coding into deep SNNs using the kernel-based dynamic threshold and dendrite to overcome the aforementioned drawback. In addition, we propose gradient-based optimization and early firing methods to further increase the efficiency of the T2FSNN. According to our results, the proposed methods can reduce inference latency and number of spikes to 22% and less than 1%, compared to those of burst coding, which is the state-of-the-art result on the CIFAR-100.

Seijoon Kim, Seongsik Park, Byunggook Na et al.

Over the past decade, deep neural networks (DNNs) have demonstrated remarkable performance in a variety of applications. As we try to solve more advanced problems, increasing demands for computing and power resources has become inevitable. Spiking neural networks (SNNs) have attracted widespread interest as the third-generation of neural networks due to their event-driven and low-powered nature. SNNs, however, are difficult to train, mainly owing to their complex dynamics of neurons and non-differentiable spike operations. Furthermore, their applications have been limited to relatively simple tasks such as image classification. In this study, we investigate the performance degradation of SNNs in a more challenging regression problem (i.e., object detection). Through our in-depth analysis, we introduce two novel methods: channel-wise normalization and signed neuron with imbalanced threshold, both of which provide fast and accurate information transmission for deep SNNs. Consequently, we present a first spiked-based object detection model, called Spiking-YOLO. Our experiments show that Spiking-YOLO achieves remarkable results that are comparable (up to 98%) to those of Tiny YOLO on non-trivial datasets, PASCAL VOC and MS COCO. Furthermore, Spiking-YOLO on a neuromorphic chip consumes approximately 280 times less energy than Tiny YOLO and converges 2.3 to 4 times faster than previous SNN conversion methods.

Seongsik Park, Seijoon Kim, Hyeokjun Choe et al.

The spiking neural networks (SNNs) are considered as one of the most promising artificial neural networks due to their energy efficient computing capability. Recently, conversion of a trained deep neural network to an SNN has improved the accuracy of deep SNNs. However, most of the previous studies have not achieved satisfactory results in terms of inference speed and energy efficiency. In this paper, we propose a fast and energy-efficient information transmission method with burst spikes and hybrid neural coding scheme in deep SNNs. Our experimental results showed the proposed methods can improve inference energy efficiency and shorten the latency.

Seongsik Park, Jaehee Jang, Seijoon Kim et al.

Memory-augmented neural networks (MANNs) are designed for question-answering tasks. It is difficult to run a MANN effectively on accelerators designed for other neural networks (NNs), in particular on mobile devices, because MANNs require recurrent data paths and various types of operations related to external memory access. We implement an accelerator for MANNs on a field-programmable gate array (FPGA) based on a data flow architecture. Inference times are also reduced by inference thresholding, which is a data-based maximum inner-product search specialized for natural language tasks. Measurements on the bAbI data show that the energy efficiency of the accelerator (FLOPS/kJ) was higher than that of an NVIDIA TITAN V GPU by a factor of about 125, increasing to 140 with inference thresholding

Seongsik Park, Seijoon Kim, Seil Lee et al.

Memory-augmented neural networks (MANNs) refer to a class of neural network models equipped with external memory (such as neural Turing machines and memory networks). These neural networks outperform conventional recurrent neural networks (RNNs) in terms of learning long-term dependency, allowing them to solve intriguing AI tasks that would otherwise be hard to address. This paper concerns the problem of quantizing MANNs. Quantization is known to be effective when we deploy deep models on embedded systems with limited resources. Furthermore, quantization can substantially reduce the energy consumption of the inference procedure. These benefits justify recent developments of quantized multi layer perceptrons, convolutional networks, and RNNs. However, no prior work has reported the successful quantization of MANNs. The in-depth analysis presented here reveals various challenges that do not appear in the quantization of the other networks. Without addressing them properly, quantized MANNs would normally suffer from excessive quantization error which leads to degraded performance. In this paper, we identify memory addressing (specifically, content-based addressing) as the main reason for the performance degradation and propose a robust quantization method for MANNs to address the challenge. In our experiments, we achieved a computation-energy gain of 22x with 8-bit fixed-point and binary quantization compared to the floating-point implementation. Measured on the bAbI dataset, the resulting model, named the quantized MANN (Q-MANN), improved the error rate by 46% and 30% with 8-bit fixed-point and binary quantization, respectively, compared to the MANN quantized using conventional techniques.

Seongsik Park, Sang-gil Lee, Hyunha Nam et al.

Nowadays deep learning is dominating the field of machine learning with state-of-the-art performance in various application areas. Recently, spiking neural networks (SNNs) have been attracting a great deal of attention, notably owning to their power efficiency, which can potentially allow us to implement a low-power deep learning engine suitable for real-time/mobile applications. However, implementing SNN-based deep learning remains challenging, especially gradient-based training of SNNs by error backpropagation. We cannot simply propagate errors through SNNs in conventional way because of the property of SNNs that process discrete data in the form of a series. Consequently, most of the previous studies employ a workaround technique, which first trains a conventional weighted-sum deep neural network and then maps the learning weights to the SNN under training, instead of training SNN parameters directly. In order to eliminate this workaround, recently proposed is a new class of SNN named deep spiking networks (DSNs), which can be trained directly (without a mapping from conventional deep networks) by error backpropagation with stochastic gradient descent. In this paper, we show that the initialization of the membrane potential on the backward path is an important step in DSN training, through diverse experiments performed under various conditions. Furthermore, we propose a simple and efficient method that can improve DSN training by controlling the initial membrane potential on the backward path. In our experiments, adopting the proposed approach allowed us to boost the performance of DSN training in terms of converging time and accuracy.

Hyeokjun Choe, Seil Lee, Hyunha Nam et al.

Near-data processing (NDP) refers to augmenting memory or storage with processing power. Despite its potential for acceleration computing and reducing power requirements, only limited progress has been made in popularizing NDP for various reasons. Recently, two major changes have occurred that have ignited renewed interest and caused a resurgence of NDP. The first is the success of machine learning (ML), which often demands a great deal of computation for training, requiring frequent transfers of big data. The second is the popularity of NAND flash-based solid-state drives (SSDs) containing multicore processors that can accommodate extra computation for data processing. In this paper, we evaluate the potential of NDP for ML using a new SSD platform that allows us to simulate instorage processing (ISP) of ML workloads. Our platform (named ISP-ML) is a full-fledged simulator of a realistic multi-channel SSD that can execute various ML algorithms using data stored in the SSD. To conduct a thorough performance analysis and an in-depth comparison with alternative techniques, we focus on a specific algorithm: stochastic gradient descent (SGD), which is the de facto standard for training differentiable models such as logistic regression and neural networks. We implement and compare three SGD variants (synchronous, Downpour, and elastic averaging) using ISP-ML, exploiting the multiple NAND channels to parallelize SGD. In addition, we compare the performance of ISP and that of conventional in-host processing, revealing the advantages of ISP. Based on the advantages and limitations identified through our experiments, we further discuss directions for future research on ISP for accelerating ML.